Analytical approaches for the. dynamics of charged particle beams

Transcription

1 Analytical approaches for the dynamics of charged particle beams J. Struckmeier, GSI Frankfurt am Main, 21 December

2 Outline Review: Hamiltonian mechanics of N particle systems Liouville s theorem Continuous description of beam dynamics Vlasov equation Moment analysis of the Vlasov equation Beam envelope equations Comparison of analytic, simulated, and measured quantities Remarks on irreversibility in computer simulations Outlook: non-liouvillean effects 2

3 The Three Pillars of today s Physics Analytical Description ("Theory") Computer Simulation Measurements 3

4 Hamilton mechanics: basis of our description of dynamical systems. Hamiltonian H for a system of N (interacting) particles H = m 2 N vi 2 + V (x 1,..., x N, t), i=1 i = 1,..., N With H given, we obtain from Hamilton s variational principle the canonical equations for each i = 1,..., N ẋ i = H v i = v i, m v i = H x i = V (x 1,..., x N, t) x i Together with the initial condition ( x1 (t 0 ), v 1 (t 0 ),..., x N (t 0 ), v N (t 0 ) ), the system of N particles is completely determined by the generally coupled system of 6N first order differential equations. 4

5 Properties of this description deterministic: The system is fully determined, no sources exist that could cause loss of information on the system s state reversible: the equations of motion ( canonical equations ) are invariant with respect to time-inversion transformations t t x i x i, v i v i. If we know the initial condition (at t 0 ) and the Hamiltonian, we can integrate the canonical equations to any instant of time, either t > t 0, or t < t 0 The canonical equations form the basis for computer simulations. In most cases, we have N sim N (representative sample). Analytically hopeless : even the three-body-problem lacks an analytical solution. For analytical studies, simplifying assumptions are necessary. 5

6 Liouville s theorem Liouville s theorem (proper) states that the volume element dγ dγ = dx 1 dv 1... dx N dv N = dx 1dv 1... dx Ndv N of a Hamiltonian N particle system is invariant with respect to canonical transformations in particular with respect to the system s time evolution. just a one invariant out of the set of Poincaré invariants. We may easily proof Liouville s theorem by showing that the determinant of the Jacobi matrix associated with a canonical transformation is unity. Note: the projection of dγ to subspaces especially to the 6-dimensional phase space or 2-dimensional phase-space planes does not necessarily provide a conserved quantity! We must be very careful if we draw conclusions that apply to subspaces! 6

7 We summarize: The single particle approach with N sim N is appropriate for computer simulations. It does not work for analytical studies. Alternative: description in terms of statistical mechanics continuous description of beam dynamics. 7

8 Continuous description of beam dynamics For the continuous description of the ion beam, we define a 6-dimensional phase-space probability density function f = f(x, v, t) f dx dv provides the probability of finding a particle inside the volume dx dv around the phase-space point (x, v) at time t. f is a smooth function of the phase-space variables. f does not provide information on individual particles. The special Liouville theorem df dt = 0 applies for systems with non-negligible Coulomb interaction only if the system can be described to sufficient accuracy by a continuous Hamilton function. This means that we must be allowed to treat the internal space charge forces analogously to an external force field. 8

9 We observe: The statistical description smoothes out the effects reflecting the actual charge granularity. The usual equation of motion of the phase-space probability density df/dt = 0 does not cover intra-beam scattering effects, for example. df/dt = 0 does not hold for stochastic cooling! 9

10 Remark on stochastic cooling Strahl Pick-up Verstaerker Kicker 10

11 The cooling process hinges on charge density fluctuations within the beam that are sampled by a pick-up electrode. This signal is used to counteract the related momentum fluctuations within the beam at some specific point further downstream the ring. The continuous description does not apply! The special Liouville theorem for f does not hold, hence stochastic cooling does not conflict with Liouville s theorem (in its special form). 11

12 Vlasov equation Explicitly, the Liouville equation for f(x, v, t) writes df dt = f 3 t + i=1 v i f x i + 3 i=1 ẍ i f v i = 0, Inserting ẍ, this equation is usually referred to as the Vlasov equation f t + v xf + 1 m (F ext + qe ssc ) v f = 0. Herein F ext stands for the external forces applied to the beam by focusing elements (such as quadrupoles) and E ssc for the smooth part of the space-charge forces. E ssc depends on f via Poisson s equation: div E ssc (x, t) = q f(x, v, t) dv. ɛ 0 12

13 We summarize: The statistical description replaces the original problem of solving 6N coupled first-order equations by one equation of motion for the 6-dimensional probability density f(x, v, t). The coupled set of Vlasov and Poisson equations provides the closed equation of motion for f if all effects due to the actual charge granularity can be neglected. Even this equation is generally not worth-while to solve directly. We do not really want to know f(x, v, t) in detail. We must switch to even more global quantities in our analytical description of ion beams. 13

14 Moment analysis of the Vlasov equation The usual way to switch to even more global physical quantities is to consider moments of f(x, v, t) x 2 i (t) = x 2 i f(x, v, t) dxdv, i = 1, 2, 3. x 2 i is proportional to the actual beam width in x i. Idea: Instead of solving the equation of motion for f, we are going to set up and solve the equations of motion for the second beam moments. The derivatives of the moments are calculated according to d x 2 dt i = x 2 f i t dxdv, inserting f/ t from the Vlasov-Poisson equation. 14

15 Integrating by parts, we obtain for each phase-space plane i = 1, 2, 3 a coupled set of moment equations d dt x 2 i 2 xi v i = 0 m d dt xi v i m v 2 i xi F ext,i q xi E ssc,i = 0 m d dt v 2 i 2 vi F ext,i 2q vi E ssc,i = 0 As usual, we define the rms emittance ε i (t) as ε 2 i (t) = x 2 i v 2 2 i xi v i The time derivative of the rms emittance may be arranged as d dt ε2 i (t) = d dt ε2 i (t) ext + d dt ε2 i (t) ssc 15

16 RMS beam envelope equations For unbunched beams with elliptic cross section in real space, we have x i E sc,i = I 4πɛ 0 cβ x 2 i x 2 + y 2. We obtain the rms envelope equation from the first two moment equations d 2 dt 2 x2 + ω 2 x(t) x 2 d 2 dt 2 y2 + ω 2 y(t) y 2 qi 4πɛ 0 mcβ qi 4πɛ 0 mcβ 1 x2 + y 2 1 x2 + y 2 ε2 x(t) x 2 3 = 0 ε2 y(t) y 2 3 = 0. We must assume the rms emittances to be approximately constant to be able to integrate this coupled set. valid for linear external forces and if no transient effects occur. 16

17 2.5 GSI QUADRUPOLE CHANNEL, SIGMA-0=120 DEG., SIGMA=35 DEG., MATCHED kv x (cm) y (cm) Cells

18 Comparison of analytic, simulation, and measured quantities In our analytic description, the second moment in x was defined as x 2 = x 2 f(x, v, t) dxdv. x2 is proportional to the actual beam width in x. In a computer simulation based on N sim representative particles, this quantity is given by x 2 = 1 N sim How do we measure this quantity? N sim i=1 x 2 i. 18

19 19

20 The result of such a slit and detector measurement is given by the current matrix I i i 1m. I =......, i n1... i nm with m the number of slit positions, and n the number of detectors. For a fixed step size x of the slit positions, the second beam moments can now be directly calculated x 2 = I 1 t m n ) 2 i kj (j x x, j=1 k=1 with I t = m n i kj, x = m n i kj j x. j=1 k=1 j=1 k=1 20

21 N Teilchensystem Bewegungsgleichungen (kan. Gleichungen) N Teilchensystem zum Zeitpunkt t0 zum Zeitpunkt t1 Projektion auf abgeleitete Größe abgeleitete Größe zum Zeitpunkt t0 Bewegungsgleichung? abgeleitete Größe zum Zeitpunkt t1 21

22 We summarize: We must clearly distinguish quantities that have only meaning at a fixed instant of time ( figures of merit, like 90 % emittances ) from quantities that are related to the system s equations of motion. If we want to understand the dynamics of measured quantities, we must ensure that these quantities are physical in the sense that they must be consistent with the equations of motion. an equation of motion must exist for the measured quantities! The rms description provides a basis to compare measured emittances, beam profiles, etc. with results of computer simulations and analytical approximations. The rms description avoids all problems that are associated with the definitions of boundaries. Best possible accuracy! 22

23 Remarks on irreversibility Consider the transformation that reverses the direction of time flow: t t x i x i, v i v i, E E, B B. We may investigate the single particle equations of motion with respect to this transformation mẍ i F ext (x i, t) q ( E i + v i B i ) = 0, i = 1,..., Nsim. We observe that this equation is indeed invariant under time reversal. We may convince ourselves that the Vlasov equation is also invariant. Earlier states are fully restored just like a movie that is reversed. The equations describe the reversible aspects of the system evolution. But: computer simulations of dynamical systems are irreversible even if the coded equations of motion are strictly reversible and the integration algorithm maintains the symplectic nature of the canonical equations ( symplectic integrator technique ). 23

24 ε(t/τ) / ε(0) Cells (t/τ) 24

25 ε(t/τ) / ε(0) Cells (t/τ) 25

26 ε(t/τ) / ε(0) Cells (t/τ) 26

27 27

28 Non Liouvillean Effects An additional equation, hence a generalization of the special Liouville theorem is necessary (Chandrasekhar (1943)): [ ] df f dt = 0 +. t If the non-liouvillean effects are small compared to the macroscopic forces (smooth space charge and external forces), we can describe them NL by the Fokker-Planck equation: [ ] f = {F i (v, t) f} + t NL v i i i,j 2 v i v j {D ij (v, t) f} D ij are referred to as elements of the diffusion tensor and the F i as elements of the drift vector that describes the dynamical friction forces. They must be determined appropriately depending on the nature of the underlying physical process. 28

29 We summarize: The actual time evolution of the simulated system always comprises irreversible aspects even if the actually coded equations of motion are strictly reversible. A computer simulation based on individual particles can never be an exact realization of a solution of the Vlasov equation. We must include these effects in our analytical description in order to fully understand our simulation results. 29